Recently, I came across a positively fascinating article:

http://www.pagetable.com/?p=39

thought others here would really enjoy this.

Thank you, that is an interesting article, and pretty much affirms what I thought was happening (at a basic level at least). I'd suspect that the combination arithmetic/shift instructions are caused by the same phenomenon... that the PLA is generating control signals for the ALU and shift registers at the same time.

That pagetable article is excellent. Somewhere I've seen a good description of the transistor-level (switch-level) design of the 6502 ALU, but I can't find it. But I did find this page (now here) with a reverse-engineered diagram of (some of) the NMOS 6502 ALU:




(Edit: I now see there's a similar diagram of this in patent 3991307 - pdf available here)

I haven't figured out the difference between the two bit-slices. (Edit: aha, it's the inversion of carry out in alternate bits, presumably to simplify the implementation.)

On the subject of implementations (and side-effects) this page goes into a little detail about how BRK, RTI, JSR and RTS share the same sequence.

There's more detail in Bill Mensch's patents - see the "description of the invention" in here for example. Surprisingly readable for a patent - even covering the design decisions.

Edit: updated image URL

Ah, Jed Margolin's website. Traipse around it, plenty of fun things. Had emailed with him for a bit concerning a game of his called Hard Drivin.

The internal emails of Atari in those years are fascinating to no end

_________________
"My biggest dream in life? Building black plywood Habitrails"